Quantitative Analysis of Liquid Fuel Mixtures with the Use of Fourier Transform Near-IR Raman Spectroscopy

Abstract

Near-infrared Fourier transform (near-IR FT) Raman spectroscopy was used to predict mass percentages of liquid fuel mixtures without the use of an internal standard. The 29 mixtures making up the calibration set were composed of varying mass percentages of unleaded gasoline, superunleaded gasoline, and diesel. Predictions were made with the use of classical least-squares with the pure components. Root mean square error (RMSE) values for all samples were 13.5, 13.8, and 5.2% (absolute error) for unleaded, superunleaded, and diesel, respectively. Using an estimate of the pure spectra determined by calibration gave RMSE values of 13.3% for unleaded, 12.2% for superunleaded, and 5.0% for diesel. A partial least-squares (PLS) model was able to partially compensate for matrix effects. Using different portions of the Raman spectra reduced the RMSE to 5.7% for unleaded, 5.2% for superunleaded, and 1.3% for diesel.